Decoding molecular switches:
from crystal structure to predictive design
Some materials can switch reversibly between two electronic states under temperature, pressure, or light. These spin-crossover (SCO) and charge-transfer-induced spin-transition (CTIST) compounds hold real promise for applications in molecular memory, sensors, and actuators. But after decades of research, a central question remains open: can we predict, from crystal structure alone, whether a new material will switch and under which conditions?
This PhD project tackles that question directly. You will combine experimental crystallography with supervised machine learning to build a predictive framework linking structural features to macroscopic switching behaviour: transition temperature, hysteresis width, abruptness. This is not a project where you apply an existing method to a standard problem. The descriptors are still being identified. The models are still being built. The database does not yet exist in the form we need. You will be constructing the scientific tools, not just using them.
On the experimental side, you will work hands-on with X-ray diffraction both on laboratory diffractometers and at large-scale synchrotron facilities such as SOLEIL, ESRF, or SPring-8. You will measure and analyse the structural response of materials to temperature, pressure, and light irradiation, and learn to extract meaningful structural information from complex diffraction data. On the computational side, you will engineer crystallographic descriptors: bond lengths, polyhedral distortion indices, packing geometry and train interpretable ML models (tree ensembles, graph neural networks) to predict the transition temperature T1/2 and hysteresis width ΔTH. Predictions will be tested against new experimental data in an iterative predict–synthesize–test loop, in collaboration with synthesis partners of the University of Tokyo.
This project sits at the intersection of condensed matter physics, structural chemistry, and data science. By the end of your PhD, you will have developed a rare combination of experimental and computational skills: crystallography, synchrotron science, data curation, and machine learning that opens doors in both academic research and the growing field of materials informatics.
Supervisors
You will be supervised by Laurent Guérin (Institut de Physique de Rennes / IRL DYNACOM, Tokyo, specialist in crystallography, synchrotron and XFEL methods, SCO/CTIST materials. You will work in close collaboration with Jean-Claude Crivello (IRL LINK, CNRS–NIMS, Tsukuba), specialist in atomistic simulations and supervised machine learning for materials science, and with Olaf Stefańczyk (University of Tokyo), specialist in the synthesis and magnetic characterization of transition-metal complexes. This ensures direct access to complementary expertise across crystallography, ML, and synthesis throughout the PhD.
Working environment
The working language is English. Your daily environment at IPR includes researchers, postdocs, and PhD students from a wide range of nationalities and scientific backgrounds. You will present your work regularly at internal and international seminars, attend conferences, and interact directly with beamline scientists during synchrotron campaigns. You will spend extended periods at the National Institute for Materials Science (NIMS) in Tsukuba and the University of Tokyo, working directly with collaboratrs on ML development and on synthesis and magnetic characterization.
